What we mean by “correct voting” is choosing the candidate under conditions of incomplete information that you would have chose had you had complete information about all of the candidates.

They’re not talking about voters opting for candidates according to their own standard of the ‘perfect candidate’. They’re talking about how the external reality of the candidate matches up to that voters impression of him. In other words, if a particular voter wants to elect the candidate who will give him the most personal financial benefit, then they are measuring whether or not the data supports that the candidate they chose is the right one for that. And if another voter wants to elect the candidate that will do the most to help the poor (rather than looking for his personal financial interest), then they’re looking at whether or not that person actually chose the person who is most likely to do that (or did the voter just think that the person seemed like he/she would do it?).

We do actually try to define which candidate is best for each voter. We call this “correct voting.”

It sounds like much of the research depends precisely on this notion of correct voting, so it’s too bad the interviewer didn’t ask more about it. I have a guess that they are assuming the correct vote is the one which is most likely to get the voter the most money, which is not necessarily how people vote, or how they should vote.